When Meta announced a $14 billion data center campus in El Paso, Texas on July 28, the headline number drew attention. What actually matters to business leaders is how Meta structured the deal — and what it signals about where AI infrastructure economics are heading.

Meta is not building this facility alone. BlackRock-managed funds will own 80 percent of the venture. Meta retains the remaining 20 percent but keeps full operational control over the campus. At financial close, Meta contributed approximately $2.3 billion in land and construction assets already underway. BlackRock contributed $4.9 billion in cash, with a further $12.5 billion in debt financing completing the capital stack.

The result: Meta maintains control of one of the largest AI compute facilities ever built in the United States while deploying a fraction of the equity it would have needed to own it outright.

What 1 Gigawatt Actually Means

The El Paso campus will deliver 1 gigawatt of compute capacity when it comes online in 2028. For context, a typical hyperscale data center might run 50 to 100 megawatts. A gigawatt-scale facility is roughly 10 to 20 times larger.

This is not expansion for expansion’s sake. The AI training runs required to push the frontier of models like Llama 6 and beyond demand compute at scales that would have seemed absurd three years ago. The models EDNA’s clients are deploying today — for AI agents, workflow automation, and voice AI employees — are trained on exactly this kind of infrastructure.

The 2028 online date is a signal too. Infrastructure of this scale takes time to permit, construct, and commission. Meta’s announcement today reflects decisions made 12 to 18 months ago. What is being announced in 2026 is the AI capacity of 2028 and beyond.

The Financial Engineering Play

The more interesting story for enterprise leaders is what Meta actually did here: it sold 80 percent of an asset it controls entirely.

This is the infrastructure equivalent of a sale-leaseback. Meta gets most of the capital back, dramatically reduces its balance sheet exposure, and frees cash for the things that generate the most return — model research, product development, and talent. BlackRock gets a hard asset with long-duration cash flows from one of the world’s most creditworthy tenants.

Private capital has been circling AI infrastructure for two years looking for exactly this kind of deal: an essential asset with a clear anchor tenant, predictable revenue, and a role in the AI buildout that is not going away. The $12.5 billion in debt financing on top of BlackRock’s equity suggests the debt markets are equally comfortable with the thesis.

Other large technology companies are watching. The capital-efficient model Meta just demonstrated — build, partially sell, retain control — is likely to become a template.

Economic Footprint

The El Paso campus will support more than 4,000 construction jobs at peak and 300 permanent operational roles once complete. Meta framed its total investment at over $10 billion, accounting for its share of the equity and its retained 20 percent stake.

Texas continues to position itself as the center of the American AI infrastructure buildout. The state’s land availability, favorable energy policy environment, and lack of state income tax make it attractive for capital-intensive, long-horizon projects exactly like this one.

What This Means for Business

For business leaders thinking about AI, this announcement has a few concrete implications.

Capacity is being built. The AI infrastructure race is not slowing down. The compute being commissioned now will underpin the models and agent capabilities of 2028 and beyond. Businesses that build AI capability today are positioning ahead of the curve, not chasing it.

AI is infrastructure. When BlackRock — a $10 trillion asset manager — structures a $14 billion deal around AI compute, it is treating AI infrastructure the same way it treats airports, toll roads, and power grids: essential, long-lived, and worth owning. That is a signal about permanence, not hype.

Cost curves will continue to fall. More compute at this scale means more competition in inference pricing and more headroom for model providers to drive down the cost of running AI workloads. The businesses that are experimenting with AI agents and automation today will benefit from meaningfully lower costs by the time this capacity comes online.

The firms that waited for AI to “stabilize” before investing in capability are going to find that the stable era began while they were still deliberating.

---

Enterprise DNA helps businesses build real AI capability across operations, data, and strategy. To explore how Omni services can accelerate your AI build-out, schedule a discovery call.